Once a sexy buzzword, artificial intelligence is rapidly morphing into something far more valuable: an everyday reality that quietly improves experiences of all sorts. That’s reflected in our roster of the most innovative companies in AI for 2022. Many of them have been at it for a while, and their efforts reflect evolution as much as revolution.
Back in 2018, for example, customer service platform LivePerson introduced its first AI-powered bots. Multiple iterations later, its “conversational AI” takes care of millions of customer engagements for companies selling everything from doughnuts to diamonds.
Then there’s Grammarly. It’s been using machine learning to parse written communications, detect errors, and make suggestions since 2009. But it’s only recently that the technology has evolved enough to let Grammarly spot truly subtle issues of readability and style, then propose sophisticated revisions to address them.
Such highly refined applications of AI share space on our list with ambitious new efforts. For instance, Eightfold AI’s Career Hub uses data to help employees find mentors and other career-boosting resources within a company. And Immunai is melding math and biology to map the human immune system, a potentially epoch-shifting step in making drug discovery more efficient.
For all of AI’s increasing ability to perform useful work, it’s also fraught with the potential to do harm, whether through deliberate misuse or due to data that’s biased or otherwise flawed. That’s why companies such as Adobe are acknowledging and addressing this uncomfortable reality. Before it ships innovations such as Photoshop’s new neural filters, it puts them through an ethics review process designed to identify—and eliminate—any problematic aspects.
Explore the full 2022 list of Fast Company’s Most Innovative Companies, 528 organizations whose efforts are reshaping their businesses, industries, and the broader culture. We’ve selected the firms making the biggest impact with their initiatives across 52 categories, including the most innovative AR/VR, design, and security companies.
For teaching bots to help human agents
LivePerson’s Conversational AI lets organizations automate straightforward customer service tasks via online chat and text messaging, so trained agents can focus on the queries that require a human touch. In 2021, the company introduced the ability to integrate Conversational AI into commerce systems, broadening its original focus on after-purchase support. Dunkin’, for example, has added QR codes to food packaging at 9,000 stores, letting customers sign up for its loyalty program by chatting with a bot. Commerce isn’t Conversational AI’s only new territory: Bella Health, a COVID-19 screening bot in use at 500 locations, is helping to detect infections before employees unwittingly spread them to coworkers. A new feature called AI Annotator has allowed support reps to improve a company’s bots on the fly, no deep knowledge of data science required; overall, there’s been a 40% year-over-year increase in automated conversations performed on LivePerson’s platform.
LivePerson is No. 21 on Fast Company’s list of the World’s 50 Most Innovative Companies of 2022.
For providing ever-smarter writing advice
Thanks to advanced AI, Grammarly has gone far beyond helping people avoid basic grammatical gaffes. The writing tool is now able to analyze text for tone, fluency, and overall readability. And rather than merely flagging problems, it now proposes full-sentence rewrites that retain the intended meaning. Grammarly has also kept up with changes in the way people use language by embracing nonbinary pronouns and neopronouns, and has begun checking for culturally stigmatizing terms such as “Chinese virus.” In 2021, the company gave its 30,000 workplace customers new features—such as brand tones and style guides—that help them apply different priorities depending on the context of a particular document. Last September, it also announced Grammarly for Developers, which lets software engineers incorporate Grammarly suggestions into any web app with just a few lines of code.
For ramping up GPT-3–and going beyond it
In May 2020, OpenAI revealed GPT-3, its landmark AI tool that can generate humanlike text and programming code. But it waited until November 2021—when it had implemented safeguards designed to prevent the technology’s use in potentially dangerous applications—before making it generally available. A month later, it began letting third-party developers tweak GPT-3’s algorithms for their own applications. Just as important, the company introduced three significant new technologies in 2021 that build on GPT-3’s advances. Codex translates natural language into computer code, opening up a possible democratization of software engineering. Dall-e turns written descriptions (“an armchair in the shape of an avocado”) into images. And Clip can identify the gist of an array of photos (say, “country line dancing”) without being trained to recognize particular subjects.
For putting Photoshop wizardry within reach
Adobe’s new neural filters use AI to bring point-and-click simplicity to visual effects that would formerly have required hours of labor and years of image-editing expertise. Using them, you can quickly change a photo subject’s expression from deadpan to cheerful. Or adjust the direction that someone is looking. Or colorize a black-and-white photo with surprising subtlety. Part of Adobe’s portfolio of “Sensei” AI technologies, the filters use an advanced form of machine learning known as generative adversarial networks. That lets them perform feats such as rendering parts of a face that weren’t initially available as you edit a portrait. Like all new Sensei features, the neural filters were approved by an Adobe ethics committee and review board that assess AI products for problems stemming from issues such as biased data. In the case of these filters, this process identified an issue with how certain hairstyles were rendered and fixed it before the filters were released to the public.
For taking the drudgery out of contracts
LinkSquares uses machine learning to automate the process of understanding contracts, eliminating the need to manually keep on top of everything from terms and conditions to legal obligations to renewal dates. By doing so, its customers can streamline risk assessments, privacy audits, and other processes that require a thorough understanding of the contractual lay of the land—all without dependence on costly outside counsel. In 2021, new integrations with Salesforce and DocuSign helped LinkSquares’s customers integrate its tools into their daily workflows. The company also bolstered its AI with new features specific to key issues such as ensuring compliance with California Consumer Privacy Act. After more than 1,000% growth in two years, LinkSquares has digested more than 3 million contracts and extracted more than 40 million data points.
6. Citrine Informatics
For speeding the development of better products
Citrine helps manufacturers of chemicals and materials reduce the time and risk involved in creating new products that are more sustainable and less dangerous. The company says that its platform can speed up R&D efforts by up to 98% compared to traditional methods that are more reliant on trial and error. In 2021, it added AI-powered features designed to forecast the performance of new ingredients and assess whether various research directions will achieve key product development goals. That helped spur a 500% year-over-year jump in weekly active users: They include a polymer manufacturer, which produced a new, lighter material for automotive use, and a specialty chemical company, which removed a hazardous ingredient from its product lineup.
7. Eightfold AI
For helping everybody to advance their careers
For too many workers, career growth is stifled by lack of access to the networking opportunities that have historically led to advancement. Eightfold’s new AI-infused Career Hub aims to level the playing field. Based on analysis of a billion careers and a million skills, it recommends mentors, projects, and internal jobs to each employee within an organization, helping to steer them in the direction they hope to go—whether that’s a promotion or a pivot into a new role. Along with launching the Career Hub, Eightfold is now working with government clients such as the state of Indiana, the country of Norway, and the U.S. Department of Labor to help upskill and reskill citizens. It also established a partnership with OneTen, the nonprofit—founded by General Catalyst Managing Director Ken Chenault and Merck Executive Chairman Ken Frazier—whose mission is to help propel the careers of 1 million Black Americans without four-year degrees over the next decade.
For securing companies by identifying the abnormal
Darktrace Immune System’s self-learning AI defends against cyberattacks by forming an understanding of an organization’s machines, processes, and people—and then springing into action when it detects signs of abnormality. One sign of its efficacy: In 2021, when the REvil ransomware gang exploited a flaw in widely used IT management software to encrypt companies’ data and demand a payment to free it, no customer using Darktrace’s technology was ensnared, the company says. Darktrace has been busy building out its capabilities to span everything from email systems to factory-floor equipment; in May 2021, it announced a partnership that brings its security measures to users of the Microsoft 365 suite, a key opportunity to rapidly increase its customer base beyond its 5,900 current subscribers. To keep ahead of threats yet to come, Cambridge, England-based Darktrace almost tripled its AI research team in 2021.
9. Scale AI
For getting AI right through better data
When the algorithms used to train machine learning models are confronted by data that falls outside typical scenarios, they often break down. Scale AI’s Scale Nucleus is a new tool designed to help companies efficiently select the best training data in the first place, rather than scramble after the fact to deal with issues such as unintended gender bias in results. One hundred customers have used Scale Nucleus to label more than 25 million pieces of data; according to Scale AI, Toyota realized a tenfold increase in annotation throughput in just weeks. The broad applicability of the company’s technology is apparent in its customer list, which includes everybody from PayPal, Pinterest, and Etsy to the U.S. Air Force.
For giving new drugs a better shot at success
One major reason why new drugs are so expensive is because so many pharmaceutical research projects fail along the way. Tel Aviv-based Immunai aims to optimize the process by using AI—and 10,000 times more data than traditional techniques—to understand the human immune system, thereby leading to more targeted therapies with higher odds of success. In 2021, the company raised a total of $275 million in Series A and Series B rounds and made two strategic acquisitions of computational biology startups. San Francisco-based Dropprint Genomics had created a database of immune cells and methods, while the Swiss company Nebion specialized in capturing and processing biological data. Immunai has also lined up some impressive partners, including Stanford, Baylor, UPenn, Sloan Kettering, and Mass General Hospital.